Repository path: workshop/translations/saadi-ghazals-radif/source-frozen.md · rendered 2026-09-09

Six Sa'di ghazals — the source's rhyming positions, frozen before any English existed

Copy-text. سعدی, «دیوان اشعار», «غزلیات», ghazals ۱۶, ۳۱, ۴۹, ۵۰, ۵۸, ۷۰, from the Ganjoor API (api.ganjoor.net/api/ganjoor/poem?url=/saadi/divan/ghazals/shN), fetched 2026-08-26 and stored verbatim at source-ganjoor-6-ghazals.txt. Sa'di died 1291/2; the text is public domain everywhere. Ganjoor is a single digital witness and this page collates nothing — no manuscript or printed edition was consulted, and variant readings are therefore unknown. This is a limit on any claim about Sa'di's practice; it is not a limit on the English, which is written against the text as given.

How a rhyming position is defined here

A ghazal rhymes […qāfiya][radif] at the second hemistich of every bayt, and at both hemistichs of the opening bayt (the مطلع). The radif is the word or phrase repeated identically at every such position; the qāfiya (here: bearer) is the word immediately before it. The radif was extracted mechanically — the longest common word-suffix over all rhyming positions, after stripping harakat and normalising ي/ی and ك/ک — by freeze.py, whose output is source-structure.json.

The six poems

Three properties of the source that the English will have to answer, recorded now

1. The depth is the poem. The shallowest of these runs is 7 positions deep. Every deep run this project has previously met was a قطعه at depth 2 to 4 (gulistan-bab1b, gulistan-bab1e), and the two it could not carry at all were depth-2 radif runs. Six poems at depth 7–8 is a different order of demand.

2. Sa'di repeats his own bearers, in four poems of six. ایطا — reusing a rhyme-word — is counted a fault by the Persian prosodists, and the ghazal that is fullest of it here (۷۰, چالاک twice in seven) is not a weak poem. 41 distinct bearers at 45 positions. This is worth freezing before any English exists because the project's standing finding about deep rhyme is that the binding constraint is the size of the English rhyme family (RS-20260823-run-depth), and it has been assumed throughout that Persian does not face that constraint. On this sample it faces a version of it.

3. Two of the six radifs are not lexical words but grammar. ۱۶'s مرا is an oblique pronoun (me), and ۷۰'s ترست is the comparative suffix plus the copula (is more —er), so that every bearer in ۷۰ is forced to be an adjective and the radif does the comparing. Four are lexical: cast, is there, arose, has fallen. A radif that is grammar and a radif that is a verb are not the same translation problem, and both are in the sample on purpose.